Pneumonia is the most common infection in the intensive care unit (ICU) (Pugh et al., 2015). Currently, gram-negative bacilli are the most common pathogenic bacteria causing ICU pneumonia (Feng et al., 2023). These bacteria have a range of resistance mechanisms against antimicrobial drugs, such as increased efflux pumps, alteration of drug binding sites, reduced permeability of the bacterial outer membrane, and production of degrading enzymes, resulting in strong antibiotic resistance (Ma et al., 2009). In recent years, the incidence of pneumonia caused by gram-negative bacilli has increased, mainly due to K. pneumoniae, A. baumannii, and P. aeruginosa (Guzek et al., 2017; Durdu et al., 2018; Solmaz and Kalın, 2021; Assefa, 2022).

Gram-negative bacilli can be isolated from approximately 67.1% of ICU patients with pneumonia, of which Escherichia coli and Pseudomonas aeruginosa account for over 80% of these cases (Sader et al., 2018). With the widespread use of broad-spectrum antibiotics, the incidence of infections with multidrug-resistant gram-negative bacilli has increased (Cillóniz et al., 2019), significantly worsening the survival rate of infected patients. Gram-negative bacilli infections have become a major challenge to global public health (Kanj et al., 2022; Dettori et al., 2023). Studies have shown that the mortality rate of gram-negative bacillary pneumonia is 14.5%-32% (Feng et al., 2019; Kofteridis et al., 2020). Patients in the ICU are often critically ill, with shorter survival times and higher mortality rates. In addition, previous studies (Casale et al., 2023; Cheema et al., 2023) showed that death in ICU patients infected with gram-negative bacillary pneumonia is related to clinical characteristics and therapeutic measures, but the predictive efficiency of various factors varies greatly, and no effective prediction model for the short-term mortality of these patients exists, nor does an evaluation of their short-term survival time. Thus, there is an important need for predictive models to determine the short-term mortality and survival analysis of ICU patients with gram-negative bacillary pneumonia.

To bridge this gap in literature, we designed this study to retrospectively analyze the case data of ICU patients with gram-negative bacillary pneumonia, identify the risk factors for 30-day mortality, establish a predictive model, stratify the risks, and assess patient survival time. The proposed model could allow clinicians to initiate early intervention and guide clinical diagnosis and treatment, thereby improving patient therapeutic outcomes.

This retrospective study revealed the predominant bacterial strains causing gram-negative bacillary pneumonia in ICU patients, with Klebsiella pneumoniae being the most common pathogen identified, followed by Acinetobacter baumannii and Pseudomonas aeruginosa. Among the 305 patients included in the study, those who succumbed within 30 days exhibited significantly higher rates of septic shock, hypoalbuminemia, chronic renal insufficiency, and acute kidney injury, along with elevated SOFA and APACHE II scores. Multifactorial logistic regression analysis identified lactic acid, tracheal intubation, and acute kidney injury as independent risk factors for 30-day mortality in these patients. Moreover, a nomogram prediction model was developed, demonstrating good discriminatory ability and providing valuable insights into risk stratification for mortality prediction in ICU patients with gram-negative bacillary pneumonia. Overall, our study provides promising insights for critical care medicine, offering a robust framework for predicting mortality risk and informing targeted interventions to optimize patient care in the management of gram-negative bacillary pneumonia in the ICU.

The incidence of pneumonia caused by gram-negative bacilli has been reported to be increasing (Assefa, 2022; Solmaz and Kalın, 2021; Guzek et al., 2017; Durdu et al., 2018; Radulescu and Gant, 2023), which is consistent with the findings of this study. ICU patients are often critically ill and immunocompromised, which makes them susceptible to opportunistic infections. Once infected with gram-negative bacilli, pneumonia can develop, increasing the patient’s mortality and medical costs (Oliveira and Reygaert, 2023). Our results indicated a mortality rate of 21.3%, which aligns with a previous report showing a 23.1% mortality rate under the same conditions (Feng et al., 2019), suggesting that patients in the ICU with gram-negative bacterial pneumonia are prone to short-term mortality, and preventing and controlling the risk factors of death is essential to improving clinical outcomes.

The gender distribution observed in our study raises interesting questions regarding the epidemiology and susceptibility of gram-negative bacillary pneumonia in ICU settings in China, particularly at our institution. This gender disparity could reflect broader trends in the prevalence of risk factors for severe infections, including lifestyle factors, genetic predispositions, or access to care, which might differ significantly between genders (Mauvais-Jarvis et al., 2020; Gay et al., 2021; Migliore et al., 2021). Previous studies have suggested that males may have a higher incidence of certain risk factors for pneumonia, such as smoking and alcohol use, which could contribute to the increased susceptibility to infections, including gram-negative bacillary pneumonia (Groeneveld et al., 2020; Corica et al., 2022). Additionally, occupational exposure to pollutants or pathogens might be higher in males, depending on regional and socioeconomic contexts (Biswas et al., 2021). It’s also possible that inherent biological differences, such as immune system function, could contribute to these disparities (Biswas et al., 2021). However, our study did not explicitly explore the reasons behind this gender distribution, and further research is needed to understand the underlying causes fully.

Logistic regression is a valuable tool for assessing disease risk factors and estimating disease occurrence probabilities under varied circumstances. Our investigation identified acute kidney injury, lactic acidosis, and tracheal intubation as independent risk factors for 30-day mortality among ICU patients infected with gram-negative bacterial pneumonia. A precedent inquiry established a nexus between renal dysfunction and mortality in 96 subjects grappling with carbapenem-resistant gram-negative bacterial pneumonia (Shen et al., 2022). Given the inflammatory nature of both pneumonia and acute kidney injury, an increased immune response could manifest in pneumonia patients concomitantly experiencing acute kidney injury, thereby complicating and exacerbating the condition while drastically diminishing survival prospects (Murugan et al., 2010; Chawla et al., 2017). Previous multivariate logistic regression analysis in pneumonia patients identified lactic acid levels as an additional independent mortality risk factor (Demirel, 2018). Gram-negative bacterial pneumonia sufferers frequently encounter multi-organ failure mainly attributable to hypoxia and insufficient perfusion. Lactic acid, a byproduct of anaerobic glucose fermentation, mirrors augmented anaerobic metabolism, serving as a direct indicator of tissue hypoxia and suboptimal perfusion, intricately linked with pneumonia severity and mortality risk (Liu et al., 2015). Furthermore, investigations have demonstrated a marked escalation in mortality risk among gram-negative bacterial pneumonia patients necessitating tracheal intubation (Lodise et al., 2021), consistent with our findings. These findings imply that individuals requiring tracheal intubation often present compromised pulmonary function, severe illness and diminished resistance to infection, rendering them particularly susceptible to mortality. Thus, to enhance the outcomes of individuals requiring tracheal intubation due to gram-negative bacterial pneumonia, comprehensive supportive care, including meticulous monitoring of pulmonary function, aggressive treatment of infections, and optimization of organ perfusion, should be prioritized, aiming to reduce mortality risks associated with their compromised condition.

The nomogram chart presents a novel predictive model characterized by high accuracy and ease of dissemination. In contrast to Feng et al. (2019), who employed multivariate logistic regression analysis on 269 cases of gram-negative bacterial pneumonia to ascertain ICU admission as a mortality risk factor, our study directly focused on ICU patients afflicted with gram-negative bacterial pneumonia. By integrating clinically relevant indicators into the nomogram chart and employing multivariate logistic regression analysis to amalgamate multiple predictive indicators, complex data are transformed into a visually accessible format, enhancing the intuitive, rigorous, and practical aspects of the prediction model for patient risk assessment. Unlike studies such as Baek et al. (2020), which solely utilized the area under the ROC curve to validate mortality predictive models in elderly patients with severe pneumonia, our research employed multiple validation methods, augmenting the credibility of the model. The Hosmer-Lemeshow test results affirm the good fit of the predictive model. With an area under the ROC curve of 0.791, our predictive model demonstrates effectiveness in forecasting the 30-day mortality risk among ICU patients with gram-negative bacterial pneumonia. Additionally, the decision curve analysis underscores a high net benefit and clinical value of the predictive model, further substantiating its utility in clinical practice.

Kaplan-Meier survival curve analysis serves as a method to depict survival time on the x-axis and survival rate on the y-axis, facilitating the assessment of certain factors’ impact on survival duration. In our investigation, we employed Kaplan-Meier survival curve analysis to evaluate the 30-day survival duration of patients categorized into low, medium, and high-risk groups based on the predictive model. The findings revealed a notable correlation between higher risk scores and shorter survival times. This stands in contrast to the study by Shen et al. (2022), which solely identified mortality risk factors in patients with gram-negative bacterial pneumonia without stratifying patients by risk or examining their short-term survival. Moreover, our study supplemented the findings with graphical representations of Kaplan-Meier survival curves, enhancing their interpretability and applicability for clinicians. Overall, upon further validation, our proposed nomogram could help guide interventions and optimize patient care in ICU settings. By stratifying patients based on their risk of mortality, healthcare providers can tailor interventions to address specific needs and mitigate adverse outcomes. For instance, patients identified as high risk according to the predictive nomogram model may benefit from closer monitoring, aggressive management of complications such as acute kidney injury, and early initiation of appropriate antimicrobial therapy targeting multidrug-resistant pathogens. Moreover, risk-based interventions can help allocate limited healthcare resources effectively, ensuring that high-risk patients receive timely and targeted interventions to improve their chances of survival. By incorporating risk stratification into clinical practice, healthcare providers can enhance the precision and effectiveness of interventions, ultimately improving patient outcomes in ICU settings.

Despite our efforts to comprehensively analyze risk factors for mortality in ICU patients with gram-negative bacillary pneumonia, several limitations should be acknowledged. Firstly, our study did not specifically assess the impact of antibiotic therapy on patient outcomes, which represents an important aspect of pneumonia management that could influence mortality rates. Second, the retrospective nature of our study design may have introduced selection bias and limited our ability to establish causal relationships between identified risk factors and mortality. Third, our study was conducted at a single center without external validation, potentially limiting the generalizability of our findings to other patient populations and healthcare settings. In addition, the possibility of unmeasured confounding variables influencing our results cannot be entirely ruled out. Lastly, the potential multicollinearity between APACHE and SOFA scores as independent variables could be considered a limitation, though the VIF coefficients were below the threshold of 5 (2.94 for APACHE and 2.09 for SOFA). Further investigation into the impact of these limitations on our findings could provide additional insights in future studies.

In conclusion, our findings reveal the important role of acute kidney injury, lactic acidosis and the need for tracheal intubation in determining the short-term mortality risk of ICU patients with gram-negative bacillary pneumonia. In regards to our proposed predictive model, this study provides important insights for devising a tool capable of early risk stratification, guiding more personalized and effective clinical interventions to improve the outcomes of these patients.

The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author/s.

The studies involving humans were approved by the Medical Ethics Committee of the First Affiliated Hospital of Fujian Medical University [2015] 084-1. The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.

Q-XL: Writing – original draft, Writing – review & editing, Conceptualization, Data curation, Formal analysis, Funding acquisition, Methodology, Supervision, Validation. ZF: Conceptualization, Data curation, Methodology, Writing – original draft, Writing – review & editing. H-CZ: Conceptualization, Data curation, Methodology, Supervision, Validation, Writing – original draft, Writing – review & editing. YZ: Conceptualization, Data curation, Methodology, Writing – review & editing. PH: Data curation, Methodology, Writing – review & editing. HL: Data curation, Writing – review & editing.